Cluster service should not access SAKAI_PRESENCE table directly
GENERAL
TESTING
GENERAL
TESTING
Description
cluster/cluster-impl/impl/src/java/org/sakaiproject/cluster/impl/SakaiClusterService.java accesses the SAKAI_PRESENCE table directly when cleaning up after servers in the cluster which have disappeared.
This is a design violation, as SAKAI_PRESENCE should only be accessed by code in the presence service. The cluster service should use an API instead.
As-is, this would cause problems if the presence implementation was replaced by something else (e.g. a non-db presence implementation).
Verified that cluster no longer refers to SAKAI_PRESENCE.
Stephen Marquard August 14, 2007 at 7:17 AM
Resolved by changes in
Stephen Marquard April 29, 2007 at 5:14 AM
r29749, r29750 add presence API and implementation for this. Attached patch is the change to cluster service to use the new API method instead of accessing the table directly.
cluster/cluster-impl/impl/src/java/org/sakaiproject/cluster/impl/SakaiClusterService.java accesses the SAKAI_PRESENCE table directly when cleaning up after servers in the cluster which have disappeared.
This is a design violation, as SAKAI_PRESENCE should only be accessed by code in the presence service. The cluster service should use an API instead.
As-is, this would cause problems if the presence implementation was replaced by something else (e.g. a non-db presence implementation).